British Telecommunications Public Limited Company

British Telecommunications Corporation

Psytechnics Limited

Michael Peter Hollier

Philip Gray

Antony William Rix

Richard J B Reynolds

Philip J Sheppard

Richard Jb Reynolds

Alexandre J Bourret

Kevin Welsby

Eleanor J Beamond

Gray Philip

Andrew Rimell

Phillip Gray

Richard John Buchan Reynolds

Philip Julian Sheppard

Richard John Reynolds

Non-intrusive speech-quality assessment uses vocal-tract models, in particular for testing telecommunications systems and equipment. This process requires reduction of the speech stream under assessment into a set of parameters that are sensitive to the types of distortion to be assessed. Once parameterized, the data is used to generate a set of physiologically-based rules for error identification, using a parametric modeling of the shape of the vocal tract itself, by comparison between derived parameters and the output of models of physiologically realistic forms for the vocal tract, and the application of physical constraints on how these can change over time.

Speech-quality assessment method and apparatus that identifies part of a signal not generated by human tract

The degree of degradation imposed on signals transmitted over a communications link from a transmitter to a receiver is tested. A predetermined test signal is transmitted using spare capacity detected by a voice activity detector when no speech is to be transmitted over the communications link. The test signal received by the receiver is compared with the reference signal, and a measure of the degradation of signals caused by the communications link is generated therefrom.

Test signalling

Two video signals, typically an original signal and a degraded version of the same signal, are analyzed firstly to identify perceptually relevant boundaries of elements forming video images depicted therein. These boundaries are then compared to determine the extent to which the properties of the boundaries defined in one image are preserved in the other, to generate an output indicative of the perceptual difference between the first and second signals. The boundaries may be defined by edges, color, luminance or texture contrasts, disparities between frames in a moving or stereoscopic image, or other means. The presence, absence, difference in clarity or difference in means of definition of the boundaries is indicative of the perceptual importance of the differences between the signals, and therefore of the extent to which any degradation of the signal will be perceived by the human viewer of the resulting degraded image. The results may also be weighted according to the perceptual importance of the image depicted&#x2014;for example the features which identify a human face, and in particular those responsible for visual speech cues.

Analysis of video signal quality

The perceptual quality of voice signals used for Voice over IP (VoIP) systems is assessed. The content of the voice data packets may be altered in order to increase the perceived quality of the VoIP system.

Method to reduce the distortion in a voice transmission over data networks

The present invention provides a test method which can determine whether a user terminal is capable of supporting a real time connection, such as a computer telephony link, with a network terminal while simultaneously supporting a connection with a WWW server.

Method to assess the quality of a voice communication over packet networks

The perceptibility of degradations caused to signals transmitted over a transmission medium is measured by generating one or more predetermined transmission degradation conditions (e) and subjecting a test signal (S) to the transmission degradation conditions (e) in a network simulation device. The degree M(e) to which the or each transmission degradation condition (e) is perceptible to the human perceptual system is measured and a data set is generated and stored for converting one or more transmission degradation conditions (e) to respective values of perceptibility M(e). The data set may be a look-up table or an empirically determined formula. The data set may then be used on live traffic (S'), by identifying objectively measured transmission degradation conditions (e) in the received signal (S&#x2032;) and retrieving from the data set in the data storage means (7) a value of perceptibility (M) associated with the transmission degradation conditions (e) so identified. In this way a measure of the subjective significance of degradation in the signal (S&#x2032;) can be derived from objective measures of degradation.

The subjective quality of an audio-visual stimulus is measured by measuring the actual synchronisation errors between the audio and visual elements of the stimulus, identifying characteristics of audio and visual cues in the stimulus, and generating a measure of subjective quality from said errors and characteristics. The nature of the cue has an effect on the perceptual significance of a given value of synchronisation error, and this can be used to relax tolerances to such errors when appropriate, or to provide an accurate measure of the quality of the signal as it would be perceived by a human subject.

Measurement of performance of communications systems

Apparatus for assessing the performance of telecommunications systems by comparison of a reference signal with the same signal as degraded by the system under test tends to give unreliable results when used to test connectionless packet-switching equipment. This is because variable delay in delivery of the individual packets can result in the test signal and reference signal becoming unsynchronised. This may lead to the apparatus recording a poorer performance than a human tester would have recorded. To overcome this problem, the invention separates the test signal into a series of individual sections, which are synchgronised and analysed separately, and a quality value, aggregated over all the sections, is derived.

Measurement of signal quality

Immersive environments for teleconferencing, collaborative shared spaces and entertainment require spatial audio. Such environments may have non-ideal sound reproduction conditions (loudspeaker positioning, listener placement or listening room geometry) where wavefront-synthesis techniques, such as ambisonics, will not give listeners the correct audio spatialization. A method disclosed for generating a sound field from a spatialized original audio signal, wherein the original signal is configured to produce an optimal sound percept at one predetermined ideal location. A plurality of output signal components are generated, each for reproduction by one of an array of loudspeakers. Antiphase output components are attenuated such that their contribution to the spatial sound percept is reduced for locations other than the predetermined ideal location. Position components defining the location of a virtual sound source, normalized to the loudspeaker distance from the ideal location, can be adapted to generate a warped sound field by raising the position components to a power greater than unity, such that the virtual sound source is perceived by listeners in the region surrounded by the loudspeakers to be spaced from the loudspeaker.

Reproduction of spatialized audio

Communications equipment is tested for perceptually relevant distortions introduced by the equipment by generating indications of the extent to which such distortion would be perceptible to a human observer, and processing high-level application data received with the input stimulus and/or generated locally relating to the intended content of the input stimulus. This allows the perceptual relevance of different distortion types to be weighted in the final output from the perceptual layer according to the nature of the signal being transmitted. The high-level information may be of a general nature, defining the type of information content in the input signal (e.g. music or speech) or may be highly defined, e.g. the input signal accompanying a video input specifying which of a limited set of objects in a virtual world is to be depicted, such that a reference copy of the image, or characteristic features of such objects can be retrieved from a store. The high-level application data may be used for other purposes, e.g. to select a coding process suitable for the nature of the information content.

Signal processing

A test stimulus includes two signals R , S. Signal S is transmitted to a customer equipment interface and over a system under test to a complementary remote apparatus. The test apparatus has two inputs processed by a number of functional elements. The first input is the reference signal R . The other input is a corresponding test signal R&prime; generated at remote test apparatus as a test signal R, identical with reference signal R , and transmitted over the system under test via the customer interface. These two versions of the same test signal are compared. Firstly, the signals R , R&prime; are processed by a psycho-acoustic model, the output of which is transmitted to a perceptual layer, and then to a conversion unit which also receives a further input from a non-optimum loudness rating unit. The output from the conversion unit is transmitted to a signal quality determination unit, which also receives an input from the customer equipment interface. The non-optimum loudness rating unit processes inputs from the customer equipment interface and signal generator. During an iterative process, the signal generator receives feedback control from the signal quality determination unit. The remote test apparatus similarly receives a modified version of signal S (viz S&prime;) from the network and compares it with a locally-generated reference version S . In an alternative arrangement the same test apparatus is connected to both test points of the network and processes both signals S and R, either serially or in parallel.

A communications system is tested to simulate a realistic conversation. A first measurement device make a call, through the system it is testing, to a second measurement device and the two devices converse using predetermined speech generated dynamically by each device in response to the signals received from the other, to simulate aspects of conversation over non-perfect communications system link. The progress of the conversation will therefore take different courses according to the quality of the link. For example, if an expected signal is detected by one device, it transmits an appropriate predetermined response to the other device. However, if the expected signal is corrupted, or not received at all, different predetermined responses (or no response at all) are transmitted to the other device. This in turn may cause the other device to repeat the original signal. Because all the responses, including those to poor quality signals, are predetermined, each device can readily determine the quality of the signals received, because each signal received by one device form the other must be one of a limited number of possible responses to the last transmission made by the one device to the other. The devices perform measurements on the sounds they receive from each other, and monitor the progress of the conversation to compute parameters describing aspects of the received signal quality.

Testing telecommunications equipment

Training process for the classification of a perceptual signal

An apparatus for improving signal quality in a communications link includes apparatus for regenerating only the speech-like characteristics of signals received over the communications link so that an estimate of the original speech signal can be retransmitted. The apparatus may include a vocal tract model coupled to a synthesizer.

Growing community of inventors

Ipswich, United Kingdom

Michael Peter Hollier